¹ Research for this paper was supported by the World Data Analysis Program of Yale University under grant #GS-614 from the National Science Foundation, and the comparative sections were done under contract #N–0014–67–A–0097–0007 from ARPA, Behavioral Sciences, monitored by the Office of Naval Research. I am grateful to Kenneth Boulding, Peter Busch, John Sullivan, and Murray Weidenbaum for comments. Of course no person or agency is responsible for errors or for the opinions expressed.

² See, for example, Isard, Walter and Ganschow, James, Awards of Prime Military Contracts by County, State, and Metropolitian Area of the United States, Fiscal Year 1960 (Philadelphia: Regional Science Research Institute, 1962)Google Scholar and Isard, Walter and Karaska, Gerald, Unclassified Defense Contracts: Awards by County, State and Metropolitian Area of the United States, Fiscal Year 1962 (Philadelphia: World Friends Research Center, 1962)Google Scholar. Weidenbaum, Murray L., “Problems of Adjustment for Defense Industries,” in Benoit, Emile and Boulding, Kenneth E. (eds.), Disarmament and the Economy (New York: Harper & Row, 1963)Google Scholar presents data on the distribution of contracts by industry. See also Clayton, James L., “Defense Spending: Key to California's Growth,” Western Political Quarterly, 15 (06 1962), 280–293 CrossRefGoogle Scholar. Science, 161 (1968), p. 448 Google Scholar reports the receipts of various universities for Department of Defense-sponsored research in the physical, biological, and social sciences. On Canada, see Rosenbluth, Gideon, The Canadian Economy and Disarmament (New York: St. Martin's Press, 1967)Google Scholar.

³ This repeated failure to shrink the military establishment back to its prewar level is a phenomenon of some interest to students of the dynamics of international arms races or Parkinson's Law. It shows up even more clearly in the data on military personnel, and goes back almost a century to demonstrate the virtual doubling of the armed forces after every war. From 1871 to 1898 the American armed forces numbered fewer than 50,000; after the Spanish-American war they never again dropped below 100,000. The aftermath of World War I saw a leveling off to about 250,000, but the World War II mobilization left 1,400,000 as the apparent permanent floor. Since the Korean War the United States military establishment has never numbered fewer than about 2,500,000 men. See Russett, Bruce M. (ed.), Economic Theories of International Politics (Chicago: Markham, 1968), p. 521 Google Scholar. Should the post-Vietnam armed forces and/or defense portion of the G.N.P. prove to be higher than in the early and mid-1960's, that will represent another diversion from private or civil public resources and a major indirect but perhaps very real “cost” of the war.

⁴ Much of the war's costs are undoubtedly absorbed by other military categories, such as the number of troops kept in other parts of the world, and reductions in new procurement and mainte-nance of old equipment not being used in the war zone. Even so, it is likely that our method understates the cost of the war to the civilian sectors of the system. President Johnson's State of the Union Message in January 1968 put the annual cost at $25 billion in 1967. For 1968 it was nearly $29 billion.

⁵ It must be made clear that this procedure is not the only plausible way of processing these data. For instance, using another theoretical model one might compute the differences between spending levels from one year to the next (t₁-t_o) and run the regression on them, giving different results. The procedure I used here assumes that it is a high level of defense spending that gets in the way of particular civilian expenditures; the alternative assumes that it is a sharp increase, regardless of level, that forces civilian cutbacks. In fact, though I report only the former, I did also process the data by the other method, finding on the whole lower r²s. In part, I hypothesize that the reason for weaker relationships with the difference model stems from varying time lags for various types of expenditure. For example, many purchases of suppliers may be sharply reduced simultaneously with the new military demands; reductions in tenured personnel, or liquidation of construction commitments, may take one or two years. To the degree that this is true, analysis of the aggregates with any particular time relationship specified (e.g., simultaneous differences, or one-year lag) would show only a weak relationship.

Also, the r²'s and regression coefficients are sensitive to the particular set of years examined. I also analyzed the World War II years and the post-war years separately, producing different and generally lower r²'s than for the entire period. Hence much of the variance found in this paper stems from differences between rather than within those two periods. Furthermore, inclusion of the war-time years with the others produces a distribution of defense percentages that would offend a purist data analyst—those for 1943 and 1944 are clear outliers nearly three standard deviations from the mean, Zand hence they exert disproportionate weight in the analysis. Excluding 1939–1945, however, would sharply reduce both the number of data points and the range of experience being analyzed. Thus I have left the earlier years in, but have also been careful to examine the scattergrams to be sure that the r²'s were not deceptively high or low because of any inordinate statistical effect of the wartime peaks.

Although I believe my handling of these two problems was appropriate, the caveat that the conclusions depend on these assumptions, and could be modified by others, is a serious one. The data are available on request from the World Data Analysis Program, 89 Trumbull Street, Yale University, New Haven, Connecticut 06520. Further analysis of the methodological problems with these data will appear in my contribution to volume 5 of Mathematical Applications in Political Science (1970).

⁶ This position is adhered to by most economists. See, for example, Benoit, and Boulding, (eds.), Disarmament and the Economy; U.S. Arms Control and Disarmament Agency, Economic Impacts of Disarmament (Washington: U.S. Government Printing Office, 1965)Google Scholar; Benoit, Emile, “The Monetary and Real Costs of National Defense,” American Economic Review, 58 (05 1968), 398–416 Google Scholar, and Isard, Walter and Schooler, Eugene, “An Economic Analysis of Local and Regional Impacts of Reduction of Military Expenditures,” Peace Research Society (International) Papers, I (1964)Google Scholar.

⁷ This index is computed by the formula

for the dependent variable. This choice of illustrative values for the G.N.P. and defense increases does not of course imply that the impact is thus in any particular build-up.

⁸ The high r² is nonetheless a bit deceptive, as a close examination of the plot discloses. If one looks at the periods of moderate defense expenditure since 1940 one finds only a mild relationship between the two variables. Most of the variance is concentrated on the differences between the moderate and high defense groups, and within the latter.

⁹ Solow, Robert M., Capital Theory and the Rate of Return (Amsterdam: North Holland, 1963)Google Scholar.

¹⁰ The costs of military procurement abroad can of course be figured more precisely and directly than we have done here, but by missing the indirect effects of inflation and diverted demand such a computation would understate the loss. Defense Department calculations of the direct cost do in fact come to but $1.5 billion; a complex independent analysis that includes the indirect effects suggests $4.0 billion, or that without the war the United States would actually have maintained a balance of payments surplus. See Dudley, Leonard and Passell, Peter, “The War in Vietnam and the United States Balance of Payments,” Review of Economics and Statistics, 50 (11 1968), 437–442 CrossRefGoogle Scholar.

¹¹ Although recent experience may make it seem obvious that public civil expenditures are likely to be inversely related to defense shares, this perception has not always been universal. Several years ago W. Glenn Campbell reported a strong inverse relationship for the 1953–63 period, in “Assuring the Primacy of National Security,” in Abshire, David M. and Allen, Richard V., National Security: Political, Military and Economic Strategies in the Decade Ahead (New York: Praeger, 1963), pp. 963–984 Google Scholar. Otto Eckstein, however, noted some questionable assumptions in Campbell's analysis and followed with a stronger expression of doubt: “I think that historical experience has been that governments are either stingy, or they're spenders. And if they're stingy about defense, they're stingy about everything. I would say that the historical record suggests that the association between civilian spending and military spending is positive, not negative.” “Discussion” in Ibid., p. 1012.

¹² The Common Defense (New York: Columbia University Press, 1961), p. 208 Google Scholar. Further masking of the impact on actual programs may stem from the inability of government agencies to reduce costs for building maintenance and tenured employees, thus forcing them in dry times to cut other expenses disproportionately.

¹³ It might be thought that this would be a case where, if there were no immediate effects of defense needs on state and local government finances, defense might nevertheless force some delayed cut-backs by the areal units. But this too is not the case, since an effort to lag local expenditures a year or two behind defense made no improvement in the fit. Apparently the federal and the areal units of government are sufficiently independent in their major revenue sources that fluctuations in the needs of the former do not seriously hamper the latter.

¹⁴ Denison, Edward F., Sources of Economic Growth in the United States and the Alternatives Before US (C.E.D., 1962)Google Scholar as cited in Mack, Ruth P., “Ecological Process in Economic Change,” American Economic Review, 58 (05, 1968), 47 Google Scholar. See also Becker, Gary S., Human Capital: A Theoretical and Empirical Analysis (New York: National Bureau of Economic Research and Columbia University Press, 1964)Google Scholar and Schultz, Theodore W., “Investment in Human Capital,” American Economic Review, 51 (03 1961), 1–17 Google Scholar. On this idea of educated manpower as a form of national capital see also the chapter by Boulding, Kenneth in Adams, Walter (ed.), The Brain Drain (New York; Macmillan, 1968)Google Scholar.

¹⁵ U.S. Bureau of the Census, Statistical Abstract of the United States, 1968 (Washington: U.S. Government Printing Office, 1968), p. 526 Google Scholar, and Statistical Abstract of the United States, 1963 (Washington: U.S. Government Printing Office, 1963), p. 544 Google Scholar. Comparable data for earlier years are not available, so this variable was not used in the computations for Table 2.

¹⁶ Brenton, Myron, “The Higher Cost of Higher Education,” New York Times Magazine, 04 28, 1968, p. 32 Google Scholar.

¹⁷ This was also the conclusion of Charles Wolf, Jr., from examining a limited amount of data for some Asian nations. See his Competition and Complementarity Between Dejense and Development—A Preliminary Approach, P-1743 (Santa Monica California: Rand Corporation, 1959)Google Scholar. Similarly, Pryor, Frederick L., in Public Expenditures in Communist and Capitalist Nations (Homewood, Ill.: Irwin, 1969)Google Scholar examined the short-run effects of defense changes on various types of expenditures in most Western developed nations for the 1950's and early 1960's. In those nations where defense spending was relatively high, typically there was an inverse relationship between defense and public civil expenditures; in the small states where the defense proportion was low, no relationship appeared. Some relationships of defense to other G.N.P. categories were found in individual countries, but there was no standard pattern.

¹⁸ Note the similar conclusion of Harold, and Sprout, Margaret, “The Dilemma of Rising Demands and Insufficient Resources,” World Politics, 20 (07 1968), 660–693 CrossRefGoogle Scholar.